2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
37 #include "opt_bootp.h"
39 #include "opt_ipstealth.h"
40 #include "opt_ipsec.h"
41 #include "opt_route.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/hhook.h>
49 #include <sys/malloc.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/socket.h>
54 #include <sys/kernel.h>
56 #include <sys/rmlock.h>
57 #include <sys/rwlock.h>
59 #include <sys/syslog.h>
60 #include <sys/sysctl.h>
63 #include <net/if_types.h>
64 #include <net/if_var.h>
65 #include <net/if_dl.h>
66 #include <net/if_private.h>
68 #include <net/route.h>
69 #include <net/route/nhop.h>
70 #include <net/netisr.h>
71 #include <net/rss_config.h>
74 #include <netinet/in.h>
75 #include <netinet/in_kdtrace.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/in_var.h>
78 #include <netinet/ip.h>
79 #include <netinet/in_fib.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet/ip_var.h>
82 #include <netinet/ip_encap.h>
83 #include <netinet/ip_fw.h>
84 #include <netinet/ip_icmp.h>
85 #include <netinet/igmp_var.h>
86 #include <netinet/ip_options.h>
87 #include <machine/in_cksum.h>
88 #include <netinet/ip_carp.h>
89 #include <netinet/in_rss.h>
91 #include <netinet/sctp_var.h>
94 #include <netipsec/ipsec_support.h>
96 #include <sys/socketvar.h>
98 #include <security/mac/mac_framework.h>
101 CTASSERT(sizeof(struct ip) == 20);
104 /* IP reassembly functions are defined in ip_reass.c. */
105 extern void ipreass_init(void);
106 extern void ipreass_vnet_init(void);
108 extern void ipreass_destroy(void);
111 VNET_DEFINE(int, rsvp_on);
113 VNET_DEFINE(int, ipforwarding);
114 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW,
115 &VNET_NAME(ipforwarding), 0,
116 "Enable IP forwarding between interfaces");
119 * Respond with an ICMP host redirect when we forward a packet out of
120 * the same interface on which it was received. See RFC 792.
122 VNET_DEFINE(int, ipsendredirects) = 1;
123 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW,
124 &VNET_NAME(ipsendredirects), 0,
125 "Enable sending IP redirects");
127 VNET_DEFINE_STATIC(bool, ip_strong_es) = false;
128 #define V_ip_strong_es VNET(ip_strong_es)
129 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es,
130 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false,
131 "Packet's IP destination address must match address on arrival interface");
133 VNET_DEFINE_STATIC(bool, ip_sav) = true;
134 #define V_ip_sav VNET(ip_sav)
135 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation,
136 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true,
137 "Drop incoming packets with source address that is a local address");
139 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */
141 static struct netisr_handler ip_nh = {
143 .nh_handler = ip_input,
144 .nh_proto = NETISR_IP,
146 .nh_m2cpuid = rss_soft_m2cpuid_v4,
147 .nh_policy = NETISR_POLICY_CPU,
148 .nh_dispatch = NETISR_DISPATCH_HYBRID,
150 .nh_policy = NETISR_POLICY_FLOW,
156 * Directly dispatched frames are currently assumed
157 * to have a flowid already calculated.
159 * It should likely have something that assert it
160 * actually has valid flow details.
162 static struct netisr_handler ip_direct_nh = {
163 .nh_name = "ip_direct",
164 .nh_handler = ip_direct_input,
165 .nh_proto = NETISR_IP_DIRECT,
166 .nh_m2cpuid = rss_soft_m2cpuid_v4,
167 .nh_policy = NETISR_POLICY_CPU,
168 .nh_dispatch = NETISR_DISPATCH_HYBRID,
172 ipproto_input_t *ip_protox[IPPROTO_MAX] = {
173 [0 ... IPPROTO_MAX - 1] = rip_input };
174 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = {
175 [0 ... IPPROTO_MAX - 1] = rip_ctlinput };
177 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */
178 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */
179 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */
181 /* Make sure it is safe to use hashinit(9) on CK_LIST. */
182 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr)));
185 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW,
186 &ip_mtu, 0, "Default MTU");
190 VNET_DEFINE(int, ipstealth);
191 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW,
192 &VNET_NAME(ipstealth), 0,
193 "IP stealth mode, no TTL decrementation on forwarding");
197 * IP statistics are stored in the "array" of counter(9)s.
199 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat);
200 VNET_PCPUSTAT_SYSINIT(ipstat);
201 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat,
202 "IP statistics (struct ipstat, netinet/ip_var.h)");
205 VNET_PCPUSTAT_SYSUNINIT(ipstat);
209 * Kernel module interface for updating ipstat. The argument is an index
210 * into ipstat treated as an array.
213 kmod_ipstat_inc(int statnum)
216 counter_u64_add(VNET(ipstat)[statnum], 1);
220 kmod_ipstat_dec(int statnum)
223 counter_u64_add(VNET(ipstat)[statnum], -1);
227 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS)
231 netisr_getqlimit(&ip_nh, &qlimit);
232 error = sysctl_handle_int(oidp, &qlimit, 0, req);
233 if (error || !req->newptr)
237 return (netisr_setqlimit(&ip_nh, qlimit));
239 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen,
240 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0,
241 sysctl_netinet_intr_queue_maxlen, "I",
242 "Maximum size of the IP input queue");
245 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS)
247 u_int64_t qdrops_long;
250 netisr_getqdrops(&ip_nh, &qdrops_long);
251 qdrops = qdrops_long;
252 error = sysctl_handle_int(oidp, &qdrops, 0, req);
253 if (error || !req->newptr)
257 netisr_clearqdrops(&ip_nh);
261 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops,
262 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE,
263 0, 0, sysctl_netinet_intr_queue_drops, "I",
264 "Number of packets dropped from the IP input queue");
268 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS)
272 netisr_getqlimit(&ip_direct_nh, &qlimit);
273 error = sysctl_handle_int(oidp, &qlimit, 0, req);
274 if (error || !req->newptr)
278 return (netisr_setqlimit(&ip_direct_nh, qlimit));
280 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen,
281 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
282 0, 0, sysctl_netinet_intr_direct_queue_maxlen,
283 "I", "Maximum size of the IP direct input queue");
286 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS)
288 u_int64_t qdrops_long;
291 netisr_getqdrops(&ip_direct_nh, &qdrops_long);
292 qdrops = qdrops_long;
293 error = sysctl_handle_int(oidp, &qdrops, 0, req);
294 if (error || !req->newptr)
298 netisr_clearqdrops(&ip_direct_nh);
302 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops,
303 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0,
304 sysctl_netinet_intr_direct_queue_drops, "I",
305 "Number of packets dropped from the IP direct input queue");
309 * IP initialization: fill in IP protocol switch table.
310 * All protocols not implemented in kernel go to raw IP protocol handler.
313 ip_vnet_init(void *arg __unused)
315 struct pfil_head_args args;
317 CK_STAILQ_INIT(&V_in_ifaddrhead);
318 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask);
320 /* Initialize IP reassembly queue. */
323 /* Initialize packet filter hooks. */
324 args.pa_version = PFIL_VERSION;
325 args.pa_flags = PFIL_IN | PFIL_OUT;
326 args.pa_type = PFIL_TYPE_IP4;
327 args.pa_headname = PFIL_INET_NAME;
328 V_inet_pfil_head = pfil_head_register(&args);
330 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET,
331 &V_ipsec_hhh_in[HHOOK_IPSEC_INET],
332 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
333 printf("%s: WARNING: unable to register input helper hook\n",
335 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET,
336 &V_ipsec_hhh_out[HHOOK_IPSEC_INET],
337 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0)
338 printf("%s: WARNING: unable to register output helper hook\n",
342 netisr_register_vnet(&ip_nh);
344 netisr_register_vnet(&ip_direct_nh);
348 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
352 ip_init(const void *unused __unused)
358 * Register statically compiled protocols, that are unlikely to
359 * ever become dynamic.
361 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL);
362 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL);
363 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL);
364 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL);
365 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL);
366 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL);
367 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL);
368 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL);
369 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL);
370 #ifdef SCTP /* XXX: has a loadable & static version */
371 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput);
374 netisr_register(&ip_nh);
376 netisr_register(&ip_direct_nh);
379 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL);
383 ip_destroy(void *unused __unused)
388 netisr_unregister_vnet(&ip_direct_nh);
390 netisr_unregister_vnet(&ip_nh);
392 pfil_head_unregister(V_inet_pfil_head);
393 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]);
395 printf("%s: WARNING: unable to deregister input helper hook "
396 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: "
397 "error %d returned\n", __func__, error);
399 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]);
401 printf("%s: WARNING: unable to deregister output helper hook "
402 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: "
403 "error %d returned\n", __func__, error);
406 /* Remove the IPv4 addresses from all interfaces. */
409 /* Make sure the IPv4 routes are gone as well. */
410 rib_flush_routes_family(AF_INET);
412 /* Destroy IP reassembly queue. */
415 /* Cleanup in_ifaddr hash table; should be empty. */
416 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask);
419 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL);
424 * IP direct input routine.
426 * This is called when reinjecting completed fragments where
427 * all of the previous checking and book-keeping has been done.
430 ip_direct_input(struct mbuf *m)
435 ip = mtod(m, struct ip *);
436 hlen = ip->ip_hl << 2;
438 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
439 if (IPSEC_ENABLED(ipv4)) {
440 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
444 IPSTAT_INC(ips_delivered);
445 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
450 * Ip input routine. Checksum and byte swap header. If fragmented
451 * try to reassemble. Process options. Pass to next level.
454 ip_input(struct mbuf *m)
456 struct ip *ip = NULL;
457 struct in_ifaddr *ia = NULL;
461 uint16_t sum, ip_len;
462 int dchg = 0; /* dest changed after fw */
463 struct in_addr odst; /* original dst address */
469 if (m->m_flags & M_FASTFWD_OURS) {
470 m->m_flags &= ~M_FASTFWD_OURS;
471 /* Set up some basics that will be used later. */
472 ip = mtod(m, struct ip *);
473 hlen = ip->ip_hl << 2;
474 ip_len = ntohs(ip->ip_len);
478 IPSTAT_INC(ips_total);
480 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip)))
483 if (m->m_len < sizeof(struct ip)) {
484 m = m_pullup(m, sizeof(struct ip));
485 if (__predict_false(m == NULL)) {
486 IPSTAT_INC(ips_toosmall);
490 ip = mtod(m, struct ip *);
492 if (__predict_false(ip->ip_v != IPVERSION)) {
493 IPSTAT_INC(ips_badvers);
497 hlen = ip->ip_hl << 2;
498 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */
499 IPSTAT_INC(ips_badhlen);
502 if (hlen > m->m_len) {
503 m = m_pullup(m, hlen);
504 if (__predict_false(m == NULL)) {
505 IPSTAT_INC(ips_badhlen);
508 ip = mtod(m, struct ip *);
511 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL);
513 /* IN_LOOPBACK must not appear on the wire - RFC1122 */
514 ifp = m->m_pkthdr.rcvif;
515 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) ||
516 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) {
517 if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
518 IPSTAT_INC(ips_badaddr);
522 /* The unspecified address can appear only as a src address - RFC1122 */
523 if (__predict_false(ntohl(ip->ip_dst.s_addr) == INADDR_ANY)) {
524 IPSTAT_INC(ips_badaddr);
528 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
529 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
531 if (hlen == sizeof(struct ip)) {
532 sum = in_cksum_hdr(ip);
534 sum = in_cksum(m, hlen);
537 if (__predict_false(sum)) {
538 IPSTAT_INC(ips_badsum);
542 ip_len = ntohs(ip->ip_len);
543 if (__predict_false(ip_len < hlen)) {
544 IPSTAT_INC(ips_badlen);
549 * Check that the amount of data in the buffers
550 * is as at least much as the IP header would have us expect.
551 * Trim mbufs if longer than we expect.
552 * Drop packet if shorter than we expect.
554 if (__predict_false(m->m_pkthdr.len < ip_len)) {
556 IPSTAT_INC(ips_tooshort);
559 if (m->m_pkthdr.len > ip_len) {
560 if (m->m_len == m->m_pkthdr.len) {
562 m->m_pkthdr.len = ip_len;
564 m_adj(m, ip_len - m->m_pkthdr.len);
568 * Try to forward the packet, but if we fail continue.
569 * ip_tryforward() may generate redirects these days.
570 * XXX the logic below falling through to normal processing
571 * if redirects are required should be revisited as well.
572 * ip_tryforward() does inbound and outbound packet firewall
573 * processing. If firewall has decided that destination becomes
574 * our local address, it sets M_FASTFWD_OURS flag. In this
575 * case skip another inbound firewall processing and update
578 if (V_ipforwarding != 0
579 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
580 && (!IPSEC_ENABLED(ipv4) ||
581 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0)
585 * ip_dooptions() was run so we can ignore the source route (or
586 * any IP options case) case for redirects in ip_tryforward().
588 if ((m = ip_tryforward(m)) == NULL)
590 if (m->m_flags & M_FASTFWD_OURS) {
591 m->m_flags &= ~M_FASTFWD_OURS;
592 ip = mtod(m, struct ip *);
597 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
599 * Bypass packet filtering for packets previously handled by IPsec.
601 if (IPSEC_ENABLED(ipv4) &&
602 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0)
607 * Run through list of hooks for input packets.
609 * NB: Beware of the destination address changing (e.g.
610 * by NAT rewriting). When this happens, tell
611 * ip_forward to do the right thing.
614 /* Jump over all PFIL processing if hooks are not active. */
615 if (!PFIL_HOOKED_IN(V_inet_pfil_head))
619 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) !=
622 if (m == NULL) /* consumed by filter */
625 ip = mtod(m, struct ip *);
626 dchg = (odst.s_addr != ip->ip_dst.s_addr);
628 if (m->m_flags & M_FASTFWD_OURS) {
629 m->m_flags &= ~M_FASTFWD_OURS;
632 if (m->m_flags & M_IP_NEXTHOP) {
633 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
635 * Directly ship the packet on. This allows
636 * forwarding packets originally destined to us
637 * to some other directly connected host.
646 * Process options and, if not destined for us,
647 * ship it on. ip_dooptions returns 1 when an
648 * error was detected (causing an icmp message
649 * to be sent and the original packet to be freed).
651 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0))
654 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no
655 * matter if it is destined to another node, or whether it is
656 * a multicast one, RSVP wants it! and prevents it from being forwarded
657 * anywhere else. Also checks if the rsvp daemon is running before
658 * grabbing the packet.
660 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on)
664 * Check our list of addresses, to see if the packet is for us.
665 * If we don't have any addresses, assume any unicast packet
666 * we receive might be for us (and let the upper layers deal
669 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) &&
670 (m->m_flags & (M_MCAST|M_BCAST)) == 0)
674 * Enable a consistency check between the destination address
675 * and the arrival interface for a unicast packet (the RFC 1122
676 * strong ES model) with a list of additional predicates:
677 * - if IP forwarding is disabled
678 * - the packet is not locally generated
679 * - the packet is not subject to 'ipfw fwd'
680 * - Interface is not running CARP. If the packet got here, we already
681 * checked it with carp_iamatch() and carp_forus().
683 strong_es = V_ip_strong_es && (V_ipforwarding == 0) &&
684 ((ifp->if_flags & IFF_LOOPBACK) == 0) &&
685 ifp->if_carp == NULL && (dchg == 0);
688 * Check for exact addresses in the hash bucket.
690 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) {
691 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr)
695 * net.inet.ip.rfc1122_strong_es: the address matches, verify
696 * that the packet arrived via the correct interface.
698 if (__predict_false(strong_es && ia->ia_ifp != ifp)) {
699 IPSTAT_INC(ips_badaddr);
704 * net.inet.ip.source_address_validation: drop incoming
705 * packets that pretend to be ours.
707 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) &&
708 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) {
709 IPSTAT_INC(ips_badaddr);
713 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
714 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len);
719 * Check for broadcast addresses.
721 * Only accept broadcast packets that arrive via the matching
722 * interface. Reception of forwarded directed broadcasts would
723 * be handled via ip_forward() and ether_output() with the loopback
724 * into the stack for SIMPLEX interfaces handled by ether_output().
726 if (ifp->if_flags & IFF_BROADCAST) {
727 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
728 if (ifa->ifa_addr->sa_family != AF_INET)
731 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr ==
733 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
734 counter_u64_add(ia->ia_ifa.ifa_ibytes,
739 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) {
740 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1);
741 counter_u64_add(ia->ia_ifa.ifa_ibytes,
749 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
751 * RFC 3927 2.7: Do not forward multicast packets from
754 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
756 * If we are acting as a multicast router, all
757 * incoming multicast packets are passed to the
758 * kernel-level multicast forwarding function.
759 * The packet is returned (relatively) intact; if
760 * ip_mforward() returns a non-zero value, the packet
761 * must be discarded, else it may be accepted below.
763 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) {
764 IPSTAT_INC(ips_cantforward);
770 * The process-level routing daemon needs to receive
771 * all multicast IGMP packets, whether or not this
772 * host belongs to their destination groups.
774 if (ip->ip_p == IPPROTO_IGMP) {
777 IPSTAT_INC(ips_forward);
780 * Assume the packet is for us, to avoid prematurely taking
781 * a lock on the in_multi hash. Protocols must perform
782 * their own filtering and update statistics accordingly.
786 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST)
788 if (ip->ip_dst.s_addr == INADDR_ANY)
790 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */
791 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) ||
792 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) {
793 IPSTAT_INC(ips_cantforward);
799 * Not for us; forward if possible and desirable.
801 if (V_ipforwarding == 0) {
802 IPSTAT_INC(ips_cantforward);
812 * IPSTEALTH: Process non-routing options only
813 * if the packet is destined for us.
815 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1))
817 #endif /* IPSTEALTH */
820 * Attempt reassembly; if it succeeds, proceed.
821 * ip_reass() will return a different mbuf.
823 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) {
824 /* XXXGL: shouldn't we save & set m_flags? */
828 ip = mtod(m, struct ip *);
829 /* Get the header length of the reassembled packet */
830 hlen = ip->ip_hl << 2;
833 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
834 if (IPSEC_ENABLED(ipv4)) {
835 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0)
841 * Switch out to protocol's input routine.
843 IPSTAT_INC(ips_delivered);
845 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p);
852 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl)
858 * The protocol slot must not be occupied by another protocol
859 * already. An index pointing to rip_input() is unused.
861 if (ip_protox[proto] == rip_input) {
862 ip_protox[proto] = input;
863 ip_ctlprotox[proto] = ctl;
870 ipproto_unregister(uint8_t proto)
875 if (ip_protox[proto] != rip_input) {
876 ip_protox[proto] = rip_input;
877 ip_ctlprotox[proto] = rip_ctlinput;
884 * Forward a packet. If some error occurs return the sender
885 * an icmp packet. Note we can't always generate a meaningful
886 * icmp message because icmp doesn't have a large enough repertoire
887 * of codes and types.
889 * If not forwarding, just drop the packet. This could be confusing
890 * if ipforwarding was zero but some routing protocol was advancing
891 * us as a gateway to somewhere. However, we must let the routing
892 * protocol deal with that.
894 * The srcrt parameter indicates whether the packet is being forwarded
895 * via a source route.
898 ip_forward(struct mbuf *m, int srcrt)
900 struct ip *ip = mtod(m, struct ip *);
901 struct in_ifaddr *ia;
903 struct sockaddr_in *sin;
907 int error, type = 0, code = 0, mtu = 0;
911 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) {
912 IPSTAT_INC(ips_cantforward);
920 ip->ip_ttl <= IPTTLDEC) {
921 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0);
925 bzero(&ro, sizeof(ro));
926 sin = (struct sockaddr_in *)&ro.ro_dst;
927 sin->sin_family = AF_INET;
928 sin->sin_len = sizeof(*sin);
929 sin->sin_addr = ip->ip_dst;
930 flowid = m->m_pkthdr.flowid;
931 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid);
932 if (ro.ro_nh != NULL) {
933 ia = ifatoia(ro.ro_nh->nh_ifa);
937 * Save the IP header and at most 8 bytes of the payload,
938 * in case we need to generate an ICMP message to the src.
940 * XXX this can be optimized a lot by saving the data in a local
941 * buffer on the stack (72 bytes at most), and only allocating the
942 * mbuf if really necessary. The vast majority of the packets
943 * are forwarded without having to send an ICMP back (either
944 * because unnecessary, or because rate limited), so we are
945 * really we are wasting a lot of work here.
947 * We don't use m_copym() because it might return a reference
948 * to a shared cluster. Both this function and ip_output()
949 * assume exclusive access to the IP header in `m', so any
950 * data in a cluster may change before we reach icmp_error().
952 mcopy = m_gethdr(M_NOWAIT, m->m_type);
953 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) {
955 * It's probably ok if the pkthdr dup fails (because
956 * the deep copy of the tag chain failed), but for now
957 * be conservative and just discard the copy since
958 * code below may some day want the tags.
964 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy));
965 mcopy->m_pkthdr.len = mcopy->m_len;
966 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t));
969 if (V_ipstealth == 0)
971 ip->ip_ttl -= IPTTLDEC;
972 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
973 if (IPSEC_ENABLED(ipv4)) {
974 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) {
975 /* mbuf consumed by IPsec */
978 if (error != EINPROGRESS)
979 IPSTAT_INC(ips_cantforward);
982 /* No IPsec processing required */
986 * If forwarding packet using same interface that it came in on,
987 * perhaps should send a redirect to sender to shortcut a hop.
988 * Only send redirect if source is sending directly to us,
989 * and if packet was not source routed (or has any options).
990 * Also, don't send redirect if forwarding using a default route
991 * or a route modified by a redirect.
994 if (!srcrt && V_ipsendredirects &&
995 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) {
996 struct nhop_object *nh;
1000 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) {
1001 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa);
1002 u_long src = ntohl(ip->ip_src.s_addr);
1004 if (nh_ia != NULL &&
1005 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) {
1006 /* Router requirements says to only send host redirects */
1007 type = ICMP_REDIRECT;
1008 code = ICMP_REDIRECT_HOST;
1009 if (nh->nh_flags & NHF_GATEWAY) {
1010 if (nh->gw_sa.sa_family == AF_INET)
1011 dest.s_addr = nh->gw4_sa.sin_addr.s_addr;
1012 else /* Do not redirect in case gw is AF_INET6 */
1015 dest.s_addr = ip->ip_dst.s_addr;
1020 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL);
1022 if (error == EMSGSIZE && ro.ro_nh)
1023 mtu = ro.ro_nh->nh_mtu;
1027 IPSTAT_INC(ips_cantforward);
1029 IPSTAT_INC(ips_forward);
1031 IPSTAT_INC(ips_redirectsent);
1042 case 0: /* forwarded, but need redirect */
1043 /* type, code set above */
1051 type = ICMP_UNREACH;
1052 code = ICMP_UNREACH_HOST;
1056 type = ICMP_UNREACH;
1057 code = ICMP_UNREACH_NEEDFRAG;
1059 * If the MTU was set before make sure we are below the
1061 * If the MTU wasn't set before use the interface mtu or
1062 * fall back to the next smaller mtu step compared to the
1063 * current packet size.
1067 mtu = min(mtu, ia->ia_ifp->if_mtu);
1070 mtu = ia->ia_ifp->if_mtu;
1072 mtu = ip_next_mtu(ntohs(ip->ip_len), 0);
1074 IPSTAT_INC(ips_cantfrag);
1078 case EACCES: /* ipfw denied packet */
1082 icmp_error(mcopy, type, code, dest.s_addr, mtu);
1085 #define CHECK_SO_CT(sp, ct) \
1086 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0)
1089 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip,
1095 if ((inp->inp_socket->so_options & SO_BINTIME) ||
1096 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) {
1097 struct bintime boottimebin, bt;
1098 struct timespec ts1;
1100 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1102 mbuf_tstmp2timespec(m, &ts1);
1103 timespec2bintime(&ts1, &bt);
1104 getboottimebin(&boottimebin);
1105 bintime_add(&bt, &boottimebin);
1109 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME,
1110 SOL_SOCKET, M_NOWAIT);
1112 mp = &(*mp)->m_next;
1116 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) {
1117 struct bintime boottimebin, bt1;
1118 struct timespec ts1;
1121 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1123 mbuf_tstmp2timespec(m, &ts1);
1124 timespec2bintime(&ts1, &bt1);
1125 getboottimebin(&boottimebin);
1126 bintime_add(&bt1, &boottimebin);
1127 bintime2timeval(&bt1, &tv);
1131 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP,
1132 SOL_SOCKET, M_NOWAIT);
1134 mp = &(*mp)->m_next;
1137 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) {
1138 struct bintime boottimebin;
1139 struct timespec ts, ts1;
1141 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1143 mbuf_tstmp2timespec(m, &ts);
1144 getboottimebin(&boottimebin);
1145 bintime2timespec(&boottimebin, &ts1);
1146 timespecadd(&ts, &ts1, &ts);
1150 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME,
1151 SOL_SOCKET, M_NOWAIT);
1153 mp = &(*mp)->m_next;
1156 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) {
1159 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1161 mbuf_tstmp2timespec(m, &ts);
1164 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC,
1165 SOL_SOCKET, M_NOWAIT);
1167 mp = &(*mp)->m_next;
1171 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR |
1173 struct sock_timestamp_info sti;
1175 bzero(&sti, sizeof(sti));
1176 sti.st_info_flags = ST_INFO_HW;
1177 if ((m->m_flags & M_TSTMP_HPREC) != 0)
1178 sti.st_info_flags |= ST_INFO_HW_HPREC;
1179 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO,
1180 SOL_SOCKET, M_NOWAIT);
1182 mp = &(*mp)->m_next;
1184 if (inp->inp_flags & INP_RECVDSTADDR) {
1185 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr),
1186 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT);
1188 mp = &(*mp)->m_next;
1190 if (inp->inp_flags & INP_RECVTTL) {
1191 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL,
1192 IPPROTO_IP, M_NOWAIT);
1194 mp = &(*mp)->m_next;
1198 * Moving these out of udp_input() made them even more broken
1199 * than they already were.
1201 /* options were tossed already */
1202 if (inp->inp_flags & INP_RECVOPTS) {
1203 *mp = sbcreatecontrol(opts_deleted_above,
1204 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT);
1206 mp = &(*mp)->m_next;
1208 /* ip_srcroute doesn't do what we want here, need to fix */
1209 if (inp->inp_flags & INP_RECVRETOPTS) {
1210 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr),
1211 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT);
1213 mp = &(*mp)->m_next;
1216 if (inp->inp_flags & INP_RECVIF) {
1219 struct sockaddr_dl sdl;
1222 struct sockaddr_dl *sdp;
1223 struct sockaddr_dl *sdl2 = &sdlbuf.sdl;
1225 if ((ifp = m->m_pkthdr.rcvif)) {
1226 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr;
1228 * Change our mind and don't try copy.
1230 if (sdp->sdl_family != AF_LINK ||
1231 sdp->sdl_len > sizeof(sdlbuf)) {
1234 bcopy(sdp, sdl2, sdp->sdl_len);
1238 offsetof(struct sockaddr_dl, sdl_data[0]);
1239 sdl2->sdl_family = AF_LINK;
1240 sdl2->sdl_index = 0;
1241 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0;
1243 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF,
1244 IPPROTO_IP, M_NOWAIT);
1246 mp = &(*mp)->m_next;
1248 if (inp->inp_flags & INP_RECVTOS) {
1249 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS,
1250 IPPROTO_IP, M_NOWAIT);
1252 mp = &(*mp)->m_next;
1255 if (inp->inp_flags2 & INP_RECVFLOWID) {
1256 uint32_t flowid, flow_type;
1258 flowid = m->m_pkthdr.flowid;
1259 flow_type = M_HASHTYPE_GET(m);
1262 * XXX should handle the failure of one or the
1263 * other - don't populate both?
1265 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID,
1266 IPPROTO_IP, M_NOWAIT);
1268 mp = &(*mp)->m_next;
1269 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t),
1270 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT);
1272 mp = &(*mp)->m_next;
1276 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) {
1277 uint32_t flowid, flow_type;
1278 uint32_t rss_bucketid;
1280 flowid = m->m_pkthdr.flowid;
1281 flow_type = M_HASHTYPE_GET(m);
1283 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) {
1284 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t),
1285 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT);
1287 mp = &(*mp)->m_next;
1294 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the
1295 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on
1296 * locking. This code remains in ip_input.c as ip_mroute.c is optionally
1299 VNET_DEFINE_STATIC(int, ip_rsvp_on);
1300 VNET_DEFINE(struct socket *, ip_rsvpd);
1302 #define V_ip_rsvp_on VNET(ip_rsvp_on)
1305 ip_rsvp_init(struct socket *so)
1308 if (V_ip_rsvpd != NULL)
1313 * This may seem silly, but we need to be sure we don't over-increment
1314 * the RSVP counter, in case something slips up.
1316 if (!V_ip_rsvp_on) {
1330 * This may seem silly, but we need to be sure we don't over-decrement
1331 * the RSVP counter, in case something slips up.
1341 rsvp_input(struct mbuf **mp, int *offp, int proto)
1348 if (rsvp_input_p) { /* call the real one if loaded */
1350 rsvp_input_p(mp, offp, proto);
1351 return (IPPROTO_DONE);
1354 /* Can still get packets with rsvp_on = 0 if there is a local member
1355 * of the group to which the RSVP packet is addressed. But in this
1356 * case we want to throw the packet away.
1361 return (IPPROTO_DONE);
1364 if (V_ip_rsvpd != NULL) {
1366 rip_input(mp, offp, proto);
1367 return (IPPROTO_DONE);
1369 /* Drop the packet */
1371 return (IPPROTO_DONE);